Considerable doubt has recently been cast on the assumption that perinatal hypoxia-ischemia is the primary cause of neonatal encephalopathy and cerebral palsy. Important epidemiological studies have suggested a wide range of alternative causal pathways, for example, stressing associations with maternal thyroid disease, or abnormalities of the hemostatic and immune systems (1, 2). A relation to inflammation has emerged: maternal fever in labor increases the offspring's risk of cerebral palsy even when there is no clinical evidence of impaired intrauterine gas exchange (3). The likelihood of specific genetic predispositions is often raised, with supporting evidence beginning to appear in some particular cases such as neonatal focal stroke.

The most consistent finding of the various epidemiological studies is that children with cerebral palsy only infrequently have evidence of perinatal hypoxia-ischemia. This should probably not surprise us. Not only can cerebral accidents presumably occur in unfortunate or predisposed individuals during the long period of intrauterine life, but the imprecise measures of fetal cerebral oxygenation and blood flow available to clinical researchers make it difficult to define perinatal hypoxia-ischemia with any certainty. Commonly quoted variables such as cardiotocography or meconium staining of the liquor are at best poor surrogate markers of brain metabolism which might be expected to predict neurologic outcome poorly (4).

So, are researchers who study the mechanisms of acute hypoxic-ischemic injury to the developing brain misguided? Not really. The most powerful epidemiological data relate to developed countries, while in the third world both neonatal encephalopathy and perinatal problems are probably more common (5). Even in the United States and Europe, accurate neuro-investigative techniques have defined at least a subgroup of infants with neonatal encephalopathy who have cerebral metabolic changes entirely characteristic of acute cerebral hypoxia-ischemia in the perinatal period (6).

Some of the first such results were obtained using magnetic resonance spectroscopy (MRS), which allows noninvasive assessment of intracellular pH (pHi), and the cerebral concentrations of ATP, phosphocreatine (PCr), inorganic phosphate (Pi), and lactate. When ATP generation is impaired energy flux is maintained by the breakdown of PCr while Pi increases, so that a decline in the ratio PCr/Pi is a precise indicator of impaired energy metabolism (7).

Experimental studies of several species of mammals using MRS and other methods have described a characteristic biphasic pattern of cerebral metabolic abnormality after cerebral hypoxia-ischemia (8–10). During experimental hypoxia-ischemia, intracerebral [PCr]/[Pi], and pHi fall, and lactate increases. Eventually [ATP] declines, but even if this transiently falls to undetectable levels, prompt resuscitation causes all these metabolites to return rapidly to normal values. However, some hours later, a second phase of metabolic abnormality begins: [PCr]/[Pi] again declines and lactate increases, although now pHi becomes alkaline. There is a dose-response relationship between the severity of the hypoxic-ischemic insult, the magnitude of the secondary changes in cerebral energy metabolism, and the extent of histologic injury.

MRS of infants with neonatal encephalopathy shows identical abnormalities in cerebral energy metabolism. The primary event cannot be observed, but cerebral energy metabolism is frequently normal soon after resuscitation, while some hours later a progressive decline in [PCr]/[Pi] and increase in pHi and lactate begins. Infants with these changes develop neurodevelopmental impairment or die, and there is a close relationship between the magnitude of the delayed disruption in energy metabolism, reduced brain growth, and the severity of neurodevelopmental impairment 1 and 4 years later (11). These findings led to the concept of “secondary energy failure,” which has been developed and extended by several groups (6, 12–14).

It has not escaped wide recognition that the apparent delay before secondary energy failure offers a rationale for developing therapies that may prevent neurologic impairment even when administered after hypoxia-ischemia. The prospect of a useful therapeutic intervention is a powerful stimulus to further investigations, and a large number of diverse interventions applied after hypoxia-ischemia have successfully reduced brain damage in many experimental systems.

The development of neural rescue therapies requires a precise understanding of the mechanisms of damage. Cells die by both apoptosis and necrosis and multiple cellular pathways are involved. Excess excitatory amino acids, changed intracellular calcium regulation, free radical generation, mitochondrial dysfunction, specific gene activation, changes in the availability of trophic factors, and the immuno-inflammatory system are all implicated. Few researchers have contributed as much to the understanding of cerebral metabolism and brain injury as Professor Siesjö, and the results presented by him and his colleagues in this issue of Pediatric Research add further to our understanding of secondary energy failure. They demonstrate in an immature animal model that the delayed phase of injury is not associated with a deficit in tissue oxygenation, but nevertheless there is impaired mitochondrial function with the production of damaging free radicals. They thus add further weight to the view that transient hypoxia-ischemia to the developing brain does not have to lead to a prolonged period of critically reduced perfusion (the “no-reflow” phenomenon) in order to cause persisting dysfunction of the energy-producing systems, in particular the mitochondria.

Mitochondrial dysfunction has far-reaching effects, leading not only to energy depletion and free radical generation, but also may directly activate caspase-9 and trigger apoptotic execution of the cell. Indeed there is evidence to suggest that mitochondrial dysfunction may persist for many months after hypoxia-ischemia (15). In children who develop neurodevelopmental impairment after perinatal hypoxia-ischemia, increased lactate and pHi can be detected in the brain for many months (16). It is not known if this is related to the very prolonged period of increased apoptotic death seen after hypoxia-ischemia in adult rats (17).

Does this increasingly clear definition of the mechanisms of hypoxic-ischemic injury conflict with epidemiological data suggesting that other causal pathways are important in the origins of neonatal encephalopathy and cerebral palsy? Probably not. The two approaches are internally consistent and scientifically valid, and look at the problem from different angles. Indeed, common ground is emerging; for example, both epidemiological and laboratory studies emphasize the importance of inflammation in brain injury. There is much to be learned from combining the two approaches. The broad vision of an epidemiological approach allied to the precise phenotypic definition available with modern neuroinvestigative techniques would be powerful indeed, especially as studies will soon have access to the huge resources of human genomics, which promises a revolution in this as in many other areas of research.